Conditions of Birth and Early Childhood Developmental Risk for Mental Disorders

The economic, health and social burden of mental illness is substantial [1], with approximately 45% of Australian adults likely to experience a mental illness in their lifetime [2], and 14% of young Australians (aged 4–17 years) estimated to have a mental illness within any 12-month period [3]. Substantive reduction of this burden could be achieved by the provision of adequate support services to those in greatest need, at the earliest opportunity [4‐6]. Growing evidence indicates the perinatal period as a critical point of contact with the health system, in which the provision of targeted supports could prevent later mental health problems and associated adversity [7].

The importance of early intervention in the first 2000 days of life—the period from conception to age 5 years—is a key focus of strategic health policy frameworks globally, including in Australia [7]. These frameworks aim to facilitate access to appropriate supports and interagency services for vulnerable families during pregnancy and the early childhood period [8, 9]. However, doing so is dependent on service providers being able to identify which children (and families) are at greatest risk for mental health problems. This has been the focus of several lines of research, with evidence for associations between multiple risk indicators evident during the perinatal period, and both externalising and internalising symptomatology emerging at age ~ 5–6 years [10‐15]. More recent research has established distinct clusters of familial and perinatal birth risk exposures that are associated with early child developmental vulnerabilities [16] as well as other adverse outcomes in later childhood (e.g., poor educational outcomes, child maltreatment, and police contact) [17, 18].

Distinct patterns of early childhood developmental vulnerability across multiple domains of function have been previously identified in a population cohort of 82,891 children who were assessed with the 2009 Australian Early Development Census (AEDC) in the state of New South Wales (NSW) [17]. The AEDC is a nationally administered triennial census of child development in the first year of full-time schooling (age ~ 5–6 years), completed by teachers who have undertaken a standardised training module. The AEDC assesses competency on five broad domains representing ‘social competence’, ‘emotional maturity’, ‘physical health and wellbeing’, ‘language and cognitive skills (school-based)’, and ‘communication skills and general knowledge’ that are underpinned by 16 subdomains. Using latent class analyses of early childhood developmental vulnerability (i.e., scores below the 10th percentile of the national distribution) on 16 subdomains of the AEDC, four risk classes were identified, including: a ‘pervasive risk’ class (4.2% of the population) characterised by a high probability of developmental vulnerability on all 16 AEDC subdomains; a ‘misconduct risk’ class (7.0%) characterised by high probability of developmental vulnerability on AEDC subdomains measuring hyperactive, inattentive and aggressive behaviour; a ‘mild generalised risk’ class (11.5%) characterised by low developmental vulnerability on most AEDC subdomains, but modest probabilities of developmental vulnerability on subdomains characteristic of academic achievement, engagement with learning, and communication, and; a ‘no risk’ class (77.0%) with virtually no developmental vulnerability on any AEDC subdomain [17, 19]. Membership in the ‘pervasive risk’ and ‘misconduct risk’ classes was associated with a threefold risk of mental disorder between the ages of 6–13 years, after accounting for socioeconomic disadvantage, notifications to child protection services, and parental mental illness [17]. All three risk classes have also been associated with an increased incidence of police contact up to age 13 years, most prominently for the ‘pervasive risk’ and ‘misconduct risk’ classes [18]. While the original LCA analyses also revealed moderate to strong associations between these risk classes and several demographic indices, perinatal events, and familial risk factors, these risk factors were not confined to information known at the time of birth.

In this large population study, two patterns of risk factors evident at the time of the child’s birth were associated with membership in distinct developmental risk classes at age ~ 5–6 years. First, exposure to several perinatal factors (e.g., grand multiparity, young maternal age, maternal smoking during pregnancy) was strongly associated with membership in all risk classes, with male sex as one of the strongest risk factors for membership in all risk classes. However, the pattern of associations observed for membership of the ‘misconduct risk’ class was somewhat different to the other risk classes, with exposures to familial factors known at the time of birth (e.g., parental mental illness, parental court charges and prenatal child protection notifications) having a generally greater magnitude of association with this class than perinatal factors. These findings could help to inform the targeted provision of preventative interventions in the earliest stages of life, to avert poor outcomes for children at risk of mental disorders [17], and other social adversities such as contact with the justice system [18]. In particular, these findings contribute to the body of research which informs the allocation of services to vulnerable families in the first 2000 days of their child’s development, before patterns of early developmental vulnerabilities are observable.

The distinct pattern of familial factors associated with the ‘misconduct risk’ class included maternal history of mental illness, maternal and paternal history of court charges, with small or no effects for perinatal factors (e.g., multiparity and grand multiparity, small for gestational age, no antenatal visit, preterm birth), and prenatal child protection exposure only maintaining association with ‘misconduct risk’ when all other factors were considered. This suggests that the behavioural profile of these children, who might carry a less adverse neurodevelopmental load than the other two classes (as suggested by minimal developmental vulnerabilities in areas characteristic of academic learning and motor function) may be more strongly determined by adverse social environments. The second pattern of risk factors associated with both the ‘pervasive risk’ and, to a lesser extent, ‘mild generalised risk’ classes was comprised of a combination of familial, perinatal and socioeconomic birth-risk factors; membership in these classes showed strong associations with perinatal indices (e.g., grand multiparity, younger mother and maternal smoking during pregnancy) and small or moderate associations with familial (e.g., maternal history of mental illness, maternal and paternal history of court charges) and socioeconomic (e.g., socioeconomic disadvantage) indices. These two patterns of risk may be detectable among expectant mothers receiving prenatal care and may thus assist to inform the earliest provision of targeted supports according to specific needs.

The NSW child protection authority delivers specific programs for vulnerable families involved with child protection services in the prenatal period [29], and Australia’s state-based health services administer voluntary universal health home visiting programs to provide support within the first two weeks of childbirth; home visits by nurses identify additional support needs and facilitate engagement with early intervention services during the postnatal period [30]. Expansion of these specialist services to identify families presenting with the factors associated with the ‘misconduct risk’ profile in childhood could potentially avert poor behavioural outcomes for the child early in development by enabling the family access to additional parenting support and suitable health services. Similarly, better access to early childhood education services for all children may be particularly beneficial to those who later show ‘pervasive risk’ and ‘mild generalised risk’ profiles of development at school entry, especially if they are also accompanied by enhanced parenting support and easy access to specialist health services. Early childhood education services are known to mitigate against poor developmental outcomes [31], including academic underachievement (which form part of the ‘pervasive risk;’ and ‘mild generalised risk’ profiles) in children experiencing adversity [32]. Indeed, a recent Australian study suggests that children with higher cumulative exposure to perinatal and sociodemographic risk factors early in life, similar to those associated with ‘pervasive risk’ and ‘mild generalised risk’ profiles, have a pattern of lower use, in early childhood education and health services [33]. Facilitating access to these programs for all vulnerable families should be a focus of health and social policy in Australia.

The current findings depart from previously reported associations between these early childhood risk classes and a range of risk factors (e.g., child protection services contact, small for gestational age, and parental mental illness) occurring before age ~ 5–6 years [19] which were associated with all risk classes to a similar degree. In the present study, with risk factors restricted to perinatal events and familial indicators known at the time of birth, we observed an emergence of two patterns of risk factors associated with either the ‘misconduct risk’ group or the ‘pervasive risk’ and ‘mild generalised risk’ groups. Moreover, the present study included additional perinatal factors that were not investigated in the earlier study, examined the effects of maternal and paternal risk factors separately, and used a more comprehensive index of child protection contacts (the earlier study was restricted to substantiated child protection reports, including out-of-home-care placements). However, the notable difference in findings for child protection contacts (which showed the strongest association for all three risk classes after male sex in the earlier study, but was associated only with the ‘misconduct risk’ class in the current study), are likely due to the lower portion of children (i.e., 0.3%) here with a prenatal child protection report known at birth compared to those (3.1%) who had been the subject of substantiated reports by age 5 years [19]. These less common prenatal child protection reports may reflect complex issues related to substance use, domestic violence, transgenerational trauma and mental ill-health, affecting the capacity of primary caregivers to adequately care for the child [29]. Finally, while previous studies have reported small to moderate associations between parental mental illness and early developmental vulnerabilities [17, 18], only maternal (but not paternal) mental illness diagnosed prior to the child’s birth was associated with membership in each of the age 5–6 year risk classes when examined simultaneously in the present study. The lack of, or smaller association with paternal mental illness is consistent with other findings in relation to externalising problems at age 5 years and emotional and behavioural development at age 3–5 years [13, 34, 35].

The main strength of this study is the large population-based sample of children and their parents, with data available during the pregnancy period [20]. The use of population-level data linkage minimises selection bias and mitigates against recall bias, while enabling the examination of rare exposures due to the large sample size. However, these findings in this study must be considered in view of several limitations. First, our index of socioeconomic disadvantage was area-based; while this is a valid approach to estimating the socioeconomic status of groups, it may be less robust as an index of individual (family) status. Second, the diagnoses of parental mental illness included only those receiving inpatient or outpatient/community-based mental health treatment, and likely underrepresents milder forms of mental illness that are treated elsewhere such as in primary care. Third, parental records were only available for children with births registered in NSW; while there is potential for this to introduce selection bias (i.e., we were unable to obtain linked parent data for children who migrated from interstate or internationally between birth and school entry), we have previously demonstrated the comparability of the sample of children with linked parental data to the entire population cohort, based on key demographic characteristics [17, 19]. Fourth, the current age of the child cohort precludes investigating the extent to which perinatal factors common to the ‘pervasive risk’ and ‘mild generalised risk’ classes may impact in later years beyond the time frame of this study [36]. For example, while there is considerable evidence supporting a link between specific perinatal factors and externalising behaviours in adolescence (e.g., low birthweight and ADHD) [37], the research examining the links between perinatal factors and internalising behaviours is mixed (e.g., preterm birth and anxiety) [38]. Fifth, some cells (i.e., grand multiparity, no antenatal visit, parental mental health and prenatal child protection contact) had relatively small sample sizes and results for these categories must therefore be interpreted cautiously. Sixth, despite male sex arising as one of the strongest risk factors for all three risk classes, this likely reflects an underrepresentation of girls as developmentally vulnerable on specific AEDC indices due to differences in the visibility of particular types of vulnerabilities; that is, externalising problems are more salient in the classroom and more commonly prevalent in boys, whereas internalising problems in children are more commonly prevalent in girls. Finally, we note that information about the ethnicity of the Australian population is not readily comparable with that recorded by other countries, since national demographic information focuses on country of birth and language background other than English, as opposed to recording identification with broad ethnic groups or race.

The present study suggests two patterns of risk factors associated with early childhood risk classes that signify vulnerability for later mental illness and other social problems. The results indicate that certain factors measurable during pregnancy and known at the time of birth could be used to inform targeted service provision for families with these patterns of risk evident in the perinatal period, shown here to be associated with distinct profiles of early childhood developmental vulnerabilities measured at school entry. These findings highlight the potential for more effective delivery of targeted services in the first 2000 days for vulnerable children and families, to improve mental health and other social outcomes in later life.